HoverFast : an accurate, high-throughput, clinically deployable nuclear segmentation tool for brightfield digital pathology images

Article Article scientifique Article HoverFast : an accurate, high-throughput, clinically deployable nuclear segmentation tool for brightfield digital pathology images eng 2024-09-26 author Liakopoulos Petros 0009-0005-2015-6795 405989 Département de médecine (DEMED) author Massonnet Julien Alfred Robert 0009-0004-9515-6100 838747 Département de pathologie et immunologie (PATIM) author Bonjour Jonatan 0009-0006-8165-6897 419099 Département de médecine (DEMED) author Mizrakli Medya Tekes author Graham Simon author Cuendet Michel 0000-0002-0754-3425 418547 Département de médecine (DEMED) author Seipel Amanda 885735 author Michielin Olivier 0000-0003-4926-6355 903944 Département de médecine (DEMED) author Merkler Doron 735735 Département de pathologie et immunologie (PATIM) author Janowczyk Andrew 405318 7022 10.21105/joss.07022 2475-9066 https://joss.theoj.org/papers/10.21105/joss.07022 In computational digital pathology, accurate nuclear segmentation of Hematoxylin and Eosin (H&E) stained whole slide images (WSIs) is a critical step for many analyses and tissue characterizations. One popular deep learning-based nuclear segmentation approach, Hover- Net (Graham et al., 2019), offers remarkably accurate results but lacks the high-throughput performance needed for clinical deployment in resource-constrained settings. Our approach, HoverFast, aims to provide fast and accurate nuclear segmentation in H&E images using knowledge distillation from HoverNet. By redesigning the tool with software engineering best practices, HoverFast introduces advanced parallel processing capabilities, efficient data han- dling, and optimized postprocessing. These improvements facilitate scalable high-throughput performance, making HoverFast more suitable for real-time analysis and application in resource- limited environments. Using a consumer grade Nvidia A5000 GPU, HoverFast showed a 21x speed improvement as compared to HoverNet; reducing mean analysis time for 40x WSIs from ~2 hours to 6 minutes while retaining a concordant mean Dice score of 0.91 against the original HoverNet output. Peak memory usage was also reduced 71% from 44.4GB, to 12.8GB, without requiring SSD-based caching. To ease adoption in research and clinical contexts, HoverFast aligns with best-practices in terms of (a) installation, and (b) containerization, while (c) providing outputs compatible with existing popular open-source image viewing tools such as QuPath (Bankhead et al., 2017). HoverFast has been made open-source and is available at andrewjanowczyk.com/open-source-tools/hoverfast. Dewey 616 Dewey 616.07 56 Département de médecine 221 137 Département de pathologie et immunologie 21F false 565 La Sclérose en plaques 908 1432 Découverte de biomarqueurs numériques pour l'oncologie de précision 1059 https://github.com/choosehappy/HoverFast https://zenodo.org/records/13838349 Journal of open source software 9 101 aa5ecf6d-0430-4fb7-b499-ce191a509e6f Article (Published version) Public Article_scientifique_from_unpaywall.pdf 2420821 application/pdf CC BY-4.0 d15b8a02e758668426c74f5cf04f1ab268d40398cd02827e163fb6682cca6957 openAccess Faculté de médecine / Section de médecine clinique / Département de médecine Faculté de médecine / Section de médecine fondamentale / Département de pathologie et immunologie

Title	Published in	Access level	OA Policy	Year	Views	Downloads

HoverFast : an accurate, high-throughput, clinically deployable nuclear segmentation tool for brightfield digital pathology images	Journal of open source software			2024	1	0

Archive ouverte UNIGE

Massonnet, Julien Alfred Robert